GPMueller
commented
5 years ago A main issue here is how to handle multi-atom basis cells. The atomistic use-case is essentially a mixture of a regular lattice (defined by Bravais vectors) and an irregular basis cell. Technically, the OVF specs forbid adding keywords which could add the necessary information.
Option 1
- Use rectangular for single-atom basis cells
- Use irregular for multi-atom basis cells
The main drawback is that the OVF specification says irregular grids should specify the positions alongside the data, which (for our purposes) unnecessarily increases the file size.
Option 2
- Use rectangular for any basis cell
The main drawback is that it is harder to check the correspondence of the geometry inside the State and the geometry inside the File Segment and an "encoding" would be required.
For example: the number of basis cell atoms could go into the Na count.
Option 3
- Extend the
OVF 2.0format or introduce anOVF 3.0orSVF 1(Spirit Vector File)- bravais vectors
- number of points per cell
- distinguish positions and values by data labels, allowing to write one segment "positions" and multiple segments with data
This would have the significant disadvantage that the files could not be read in by micromagnetic codes which strictly adhere to the OVF specs.
Questions
- which checks are necessary when reading in a file?
- number of spins? -> in both directions, a warning might be enough
- number of cells? -> warning?
- spatial dimensions?
- which is more important: compatibility or proper representation of the atomistic lattice?
The
xbaseetc parameters are definitely being written incorrectly. Probably there are more problems, see https://github.com/spirit-code/ovf#file-format-v20-specification-